39 results about "Auto segmentation" patented technology

Systems and methods for automated segmentation of anatomical structures, such as the human heart. The systems and methods employ convolutional neural networks (CNNs) to autonomously segment various parts of an anatomical structure represented by image data, such as 3D MRI data. The convolutional neural network utilizes two paths, a contracting path which includes convolution / pooling layers, and an expanding path which includes upsampling / convolution layers. The loss function used to validate the CNN model may specifically account for missing data, which allows for use of a larger training set. The CNN model may utilize multi-dimensional kernels (e.g., 2D, 3D, 4D, 6D), and may include various channels which encode spatial data, time data, flow data, etc. The systems and methods of the present disclosure also utilize CNNs to provide automated detection and display of landmarks in images of anatomical structures.

A system that automatically analyzes search queries made by visitors on search engines in order to automatically segment search queries and visitors in order to make the advertisements displayed by the search engines more targeted and so more valuable for the advertiser, and allow the search engine to increase the revenue related to advertisement sales and the advertiser to be more profitable.

The present invention relates to an automatic segmentation method for a breast MRI focus based on inter-frame correlation. The method comprises: reading an MRI image; preprocessing the image, and performing coarse segmentation on the preprocessed image to determine an initial contour of a focus; and by adopting an improved C- V level set model method, performing fine segmentation on a coarsely segmented image that is obtained in the prior step, further refining the tumor contour on the basis of a coarsely segmented contour, and optimizing an obtained finely segmented result by combining inter-frame correlation. The method provided by the present invention has higher accuracy and can better segment the focus.

A system and method for enhanced viewing of rib metastasis in CT volume data is disclosed. The system and method receive input CT volume data and display slices of the CT volume data. Ribs are automatically segmented from the CT volume data, ordered and labeled. A 3D visualization of the ribs is generated and displayed. Alterations in the rib structure is automatically detected using shape based analysis of the ribs. The alterations are marked as candidate locations for rib metastasis in the displayed slices and 3D visualization in order to assist in the diagnosis of rib metastasis.

There is provided a method for automatically segmenting lung nodules in a three-dimensional (3D) Computed Tomography (CT) volume dataset. An input is received corresponding to a user-selected point near a boundary of a nodule. A model is constructed of the nodule from the user-selected point, the model being a deformable circle having a set of parameters β that represent a shape of the nodule. Continuous parts of the boundary and discontinuities of the boundary are estimated until the set of parameters β converges, using dynamic programming and Expectation Maximization (EM). The nodule is segmented, based on estimates of the continuous parts of the boundary and the discontinuities of the boundary.

The invention discloses an automatic segmentation method for a lesion area in a digital pathological full slice image. The method comprises steps that an offline training phase and an online prediction phase are included, for the training phase, a digital pathological full slice image in a full slice database with the labeled lesion area is sampled to obtain a large number of labeled image blocks,and a classifier is trained; for the prediction stage, an image block matrix is obtained through uniformly sampling an unknown digital pathological full slice image, multiple probability matrixes areobtained by the classifier, the probability matrixes are processed and binarized, and the obtained contour is mapped back to the unknown digital pathological full slice image to obtain the segmentation result. The method is advantaged in that automatic segmentation of the lesion area of the unknown full slice can be achieved simply through the full-slice database with the labeled lesion area, notonly lesion categories of all the possible lesion areas in the full slice are shown, but also the contour and position distribution of each lesion area are further accurately displayed, and comprehensive, intuitive and accurate prediction and analysis of the unknown full-slice lesion status are realized.

Disclosed herein are techniques for enhancing the accuracy of atlas-based auto-segmentation (ABAS) using an automated structure classifier that was trained using a machine learning algorithm. Also disclosed is a technique for training the automated structure classifier using atlas data applied to the machine learning algorithm.

The invention discloses a breast tumor ultrasonoscopy automatic segmentation method based on mean shift and divide. A breast tumor ultrasonoscopy is filtered by means of a pyramid mean shift algorithm, the filtered breast tumor ultrasonoscopy is segmented by means of a divide algorithm, the minimum grey level of a specific area of interest in the breast tumor ultrasonoscopy which is segmented through the divide algorithm is calculated according to the experiential knowledge that tumors are generally located on the middle portion or the upper portion of the ultrasonoscopy and the average image intensity is low, the ultrasonoscopy which is segmented through the divide algorithm is traversed, a pixel is regarded as a foreground if the grey level of the pixel is equal to the minimum grey level, the pixel is regarded as a background if the grey level of the pixel is not equal to the minimum grey level, and thus a target tumor area is obtained, namely a final binary image of the tumor segmented result. By means of the breast tumor ultrasonoscopy automatic segmentation method based on mean shift and divide, the boundary of a tumor in the breast tumor ultrasonoscopy is clear and can be retrieved automatically, and the breast tumor ultrasonoscopy can be segmented fast, accurately and automatically.

Systems and methods described herein employ multiple phase-contrast images with various relative phase shifts between light diffracted by a sample and light not diffracted by the sample to produce a quantitative phase image. The produced quantitative phase image may have sufficient contrast for label-free auto-segmentation of cell bodies and nuclei.

A manual windowing with auto-segmentation assistance method and system are disclosed that include defining manual tags to be assigned to one or more windows via a tag selection module, assigning the defined manual tags to the one or more windows via a manual tag assignment module and auto-segmenting the one or more windows within the display medium by producing auto-segmentation tags via the auto-segmentation module. Furthermore the method and system according to this invention provide for mapping tags to the auto-segmentation tags via a tag mapping module and merging the manual tags and the auto-segmentation tags via a merging module.

The invention discloses a liver CT automatic segmentation method based on deep shape learning, and the method comprises the steps: firstly building a liver segmentation data set, carrying out the preprocessing, and carrying out the coarse segmentation of a liver CT through the liver segmentation; secondly, establishing a liver shape set, learning a liver shape by using a variational auto-encoder, constructing a geometrical shape regularization module, and then adding the geometrical shape regularization module into liver segmentation to obtain a liver segmentation model constrained by geometrical shape consistency for automatic segmentation of liver CT. According to the method, the expressed shape features are creatively added into the existing deep segmentation network through the regularization module, and shape prior information is introduced in the training process of the convolutional neural network, so that the regularity and generalization ability of the segmentation model can be improved, and the segmentation result is enabled to better conform to the medical anatomy characteristics of the standard liver. The method has the advantages of being automatic, high in precision and capable of being migrated and expanded, and automatic and accurate segmentation of the abdominal large organs, such as the liver, can be achieved.

Content items may be segmented and labeled by topic to provide for the capture, analysis, indexing, retrieval and / or distribution of information within information rich media, such as audio or video, with greater functionality, accuracy and speed. The segments and other related information may be stored in a database and made accessible to users through, for example, a search service and / or an on-demand service. Automatic segmentation may include receiving a text representation, calculating relevance intervals based on the text representation, determining a nodal representation based on the relevance intervals, and determining segments of the content item based on the nodal representation.

The invention relates to a fully automatic lumbar vertebral image segmentation method based on pre-emphasis strategy, The method comprises the following steps: a data generation method based on humanlumbar vertebrae structure and magnetic resonance contrast characteristics automatically generates a large number of spine magnetic resonance images with rich structural diversity and texture diversity of the spine, and completes the training of the lumbar vertebrae image segmentation model; Using the trained segmentation model, the automatic segmentation of vertebral body and intervertebral discin spinal MRI data is realized. The invention can solve the problem of the limitation of the data of the traditional training model, and has high model generalization ability. There are many kinds oflumbar magnetic resonance images which are adaptable to different hospitals, different scanning machines and different scanning parameters.

Provided herein are systems and methods for automatically segmenting a 3D model of a patient's teeth. A patient's dentition may be scanned. The scan data may be converted into a 3D model, including a graph-based representation of the 3D model. The graph-based representation can be input into a machine learning model to train the machine learning model to segment the 3D model into individual dental components. Trained machine learning models can also be used to segment graph-based representations of a 3D model of a patient's teeth.

The invention belongs to the field of medical imaging, in particular to a CT image segmentation method. The method comprises the steps that 1, cross section position images before and after thyroid gland enhancement and registering is conducted; 2, a thyroid gland tissue template extracted; 3, thyroid tissues are segmented. The method has the advantages of high speed, accurate and the like, and the technical demands of thyroid gland CT image automatic diagnosis can be met.

The system for determining the three-dimensional functional liver segment based on the medical image comprises an image input module, an image classification module, a functional image resolution enhancement module, an anatomical image liver segment automatic segmentation module, a three-dimensional functional liver segment acquisition module and a result output module. The functional influence resolution enhancement module performs resolution enhancement processing on the functional image to obtain a high-resolution enhanced functional image, and the anatomical image liver segment automatic segmentation module performs liver segment automatic segmentation on the anatomical image to obtain a refined three-dimensional liver segment anatomical image. A functional image and an anatomical image are input, the input images are classified, resolution enhancement processing is performed on the functional image, three-dimensional refining processing and registration fusion are performed on the anatomical image to obtain a three-dimensional functional liver segment, and an image result is output and imported into a radiotherapy plan design system for radiotherapy dose calculation. According to the invention, three-dimensional functional liver segment judgment can be realized, the method is used for radiotherapy, the segmentation precision is improved, and the treatment cost of a patient is reduced.

The invention discloses a cerebrovascular atlas construction method. The method comprises preprocessing and labeling the MRA data; screening required patches through an algorithm in the data segmentation process; inputting the data into a variant U-Net network for training and prediction; converting a data set of obtained prediction results into the same coordinate system through registration, calculating the similarity, and allocating the data set to a corresponding feature space, and constructing a cerebrovascular atlas. Blood vessels can be automatically segmented only by preprocessing thenew MRA blood vessel sample data and carrying out coordinate conversion on the atlas. According to the method, the calculated amount is greatly reduced in data segmentation, the blood vessel atlas making method is provided, the subjectivity of manual operation of a traditional method is solved, and rapid, simple, accurate and automatic segmentation of cerebral vessels is achieved.

A color image segmentation method is provided. The color image segmentation method includes the steps of: (a) calculating a predetermined value representing the degree of difference form the color of peripheral pixels by using pixel values of an input image; (b) obtaining a converted image by converting a calculated value into a value of a predetermined scale; and (c) segmenting the converted image. According to the color image segmentation method, a robust and an automatic segmentation is possible, and a segmentation speed is high even when segmenting an image containing much noise.

The invention discloses a fiber identification and segmentation method. The fiber identification and segmentation method can automatically segment a single curve in a picture. Compared with manual segmentation, the fiber identification and segmentation method disclosed by the invention greatly improves efficiency, and a curve head slope, a tail slope, length and an interval variance are used for determining the rough contour of a fiber edge so as to realize the single segmentation of the fiber.

The invention relates to an abdominal muscle labeling method and a labeling device based on a plurality of sub-region templates. According to the characteristics of wide abdominal muscle area, large shape change and diversity, the abdominal muscle is divided into four sub-regions according to the shape characteristics and anatomical significance of the abdominal muscle. Collect standard annotationdata to construct standard abdominal muscle annotation database. At the same time, according to the standard annotation database, according to the four sub-regions of abdominal muscle, multiple shapetemplates of each sub-region are constructed to form shape template library. Based on the annotation database and shape template database, the four sub-regions of abdominal muscle are automatically segmented and annotated by multi-template matching method. At the same time, the marking method and the marking device of the invention can view and revise the result, and refine and revise the markingresult. The automatic labeling method of the invention greatly reduces the daily work burden of the doctor and ensures the accuracy of the medical data labeling.

The present invention includes a composition, method, method of making, and a kit for using an enteric contrast agent formulation comprising an enteric contrast medium comprising particles comprising atoms of an element with an atomic number from 70 to 77, and a pharmaceutically acceptable vehicle in which the particles are dispersed.

The invention provides a pancreas segmentation method based on a deep convolutional neural network. The pancreas segmentation method comprises the following steps: acquiring a computed tomography image at the pancreas of a patient; preprocessing a computed tomography image, then inputting the preprocessedcomputed tomography image into the trained deep convolutional neural network model, and enabling the trained deep convolutional neural network model to carry out automatic segmentation and obtain a preliminary pancreas segmentation result; carrying out implicit contour simulation on the preliminary pancreas segmentation result through a distance regularization level set, determining a final pancreas boundary through an optimization algorithm, and obtaining a final pancreas segmentation result after noise reduction and normalization processing are conducted; sending a region of interest into three neural networks for preliminary segmentation, performing data enhancement on a two-dimensional pancreas image to obtain enough training and verification data, and obtaining a preliminary segmentation result of a pancreas region, so that a segmentation boundary of a pancreas is clear and smooth, and prior knowledge of anatomy can be simply and conveniently added into a segmentation model.

The invention relates to a method for synchronously segmenting multiple organs of a lung. The method comprises the following steps: a, performing interactive data annotation on the multiple organs of the lung; b, performing data preprocessing on the data annotation; c, constructing a data set for the data preprocessing result; d, enhancing and expanding the data set; e, carrying out DC-U-Net network model training on the data set to obtain a DC-U-Net training model; f, carrying out automatic segmentation on multiple organs of the lung based on the DC-U-Net training model. According to the invention, lung parenchyma, blood vessel, bronchus and pulmonary nodule areas can be automatically extracted based on the deep learning technology, and synchronous segmentation is realized.

Disclosed is a computer-implemented method which encompasses registering a tracked imaging device such as a microscope having a known viewing direction and an atlas to a patient space so that a transformation can be established between the atlas space and the reference system for defining positions in images of an anatomical structure of the patient. Labels are associated with certain constituents of the images and are input into a learning algorithm such as a machine learning algorithm, for example a convolutional neural network, together with the medical images and an anatomical vector and for example also the atlas to train the learning algorithm for automatic segmentation of patient images generated with the tracked imaging device. The trained learning algorithm then allows for efficient segmentation and / or labelling of patient images without having to register the patient images to the atlas each time, thereby saving on computational effort.

The invention discloses a brain age prediction method based on automatic identification of fiber bundles, and aims to solve the problems that fiber bundles sensitive to age change cannot be accurately positioned and fiber characteristics are averaged in an existing method for predicting brain age by extracting white matter characteristics based on voxels. A whole-brain fiber bundle graph is obtained from a diffusion tensor imaging image through a fiber tracking algorithm, fiber bundles with anatomical significance are automatically segmented and recognized according to a white matter graph of the brain, diffusion indexes are quantified along the fiber bundles, finally, the diffusion indexes serve as features to be input into a brain age prediction model, and the accuracy of the prediction model is tested. According to the method, tensor values are extracted through a fiber bundle automatic identification method to serve as white matter features, subtle changes of a white matter fiber bundle microstructure in the aging process can be better reflected by constructing a brain age prediction model, and fiber bundles with sensitive changes in the aging process are found out.

Disclosed is a computer-implemented method which encompasses registering a tracked imaging device such as a microscope having a known viewing direction and an atlas to a patient space so that a transformation can be established between the atlas space and the reference system for defining positions in images of an anatomical structure of the patient. Labels are associated with certain constituents of the images and are input into a learning algorithm such as a machine learning algorithm, for example a convolutional neural network,together with the medical images and an anatomical vector and for example also the atlas to train the learning algorithm for automatic segmentation of patient images generated with the tracked imaging device. The trained learning algorithm then allows for efficient segmentation and / or labelling of patient images without having to register the patient images to the atlas each time, thereby saving on computational effort.